On 9/8/13 4:12 PM, Nico Kadel-Garcia wrote: > This is why you either update a system version that is high enough, or you use > the "get-deps.sh" to pull down versions locally and compile them statically. > I've been publishing up-to-date get-deps.sh, slightly more consistent in > layout > and up-to-date, at > https://github.com/nkadel/subversion-1.8.x-srpm/blob/master/get-deps.sh.
He does have a legitimate complaint, the problem is that Debian's serf isn't new enough and our configure should be complaining. > Can we possibly get that updated get-deps.sh into the main code line? Make a patch and submit it along with a commit message explaining the reasoning for your changes. Some comments from looking at your script now: 1) The nested command substitutions in setting the SQLITE need to be broken up to be compatible with POSIX shells. You should avoid using $() since there are some shells that don't implement this (despite the fact that it's in the POSIX standard). 2) What's the motivation for reordering stuff all over the place. It just seem gratutitous and it doesn't really matter what order get-deps.sh downloads things. What it does achieve is to make your differences very difficult to actually see. 3) As far as I know the newest version of APR iconv is 1.2.1, your script is defaulting to 1.2.6 4) Is there some particular reason to want SQLITE 3.7.17 versus 3.7.15.1? We don't typically bump version numbers once a 1.x.0 has been released unless there's a specific reason to. 5) There's no reason to download gtest since nothing actually uses it yet. 6) What's the point of all the || return 1? The existing test when using user provided deps list is to show the usage if someone tries to use a deps name we don't have implemented. I mean I can see the value in checking that the commands actually suceeded but when you just run the script with no command it doesn't do anything with that return and when there is an error it just prints the usage, which is hugely confusing.
